[Pharo-project] [Metacello] Metacello tests failing in 1.2

Dale Henrichs dhenrich at vmware.com
Thu Feb 24 15:15:12 CET 2011

I haven't released my changes yet ... once I'm satisfied with the the behavior on PharoCore1.2 (that's where I'm doing development) I now have to get the new version of Gofer ported to Squeak and GemStone before I can make the full release ... otherwise the tests will be failing on Squeak/GemStone ...

If the goal is to make the tests green on Pharo1.2, you might just add #expectedFailures as non-metacello extension methods to the troublesome tests ... I think that this is the exact use case that #expectedFailures is aimed at ... the tests will be green for Pharo1.2 and I'll be able to take the necessary time to propogate the necessary changes...

On Feb 24, 2011, at 5:50 AM, laurent laffont wrote:

On Thu, Feb 24, 2011 at 2:31 PM, Dale Henrichs <dhenrich at vmware.com<mailto:dhenrich at vmware.com>> wrote:

I've actually made a pass and cleaned up the tests ... you can compare with my work:)... I need to write some more tests to make sure that Metacello is still behaving correctly, I think I've added enough coverage, but I need to look a little closer ...

My guess is that you didn't reset the resource ... the resource structure is cached, so when you make changes to the code that is used to create the resource, you need to reset it as well...

Yes that was that !  Still 9 errors but as you've already done it ..... your changes has been integrated in PharoCore 1.2 ?



On Feb 24, 2011, at 12:10 AM, laurent laffont wrote:

On Thu, Feb 17, 2011 at 12:33 AM, Dale Henrichs <dhenrich at vmware.com<mailto:dhenrich at vmware.com><mailto:dhenrich at vmware.com<mailto:dhenrich at vmware.com>>> wrote:
On 02/15/2011 05:46 PM, Guillermo Polito wrote:
Hi Dale!

There are 13 failing tests on the 1.2 dev build since a week ago (
).  Do we have the right version?  Do you have any clue so I don't dive
in the code blindly? :)


Ahhhh, It looks the Gofer changes for Issue 3660 have been integrated, which means that the sorting order for mcz files has been changed, which means that the Metacello tests that are sensitive (or validating) the sort order have broken ...

It will be at least a couple of days before I can get these issues fixed I have to propogate the gofer fixes to GemStone and Squeak... the normal operation of Metacello shouldn't be affected by these failing tests ... the test cases involved include examples of mcz files with "incorrect" branch naming scheme...

I've created Metacello Issue 108: "Need to update Metacello (tests and Gofer config) to integrate changes from Pharo Issue 3660"[1] to track progress.

[1] http://code.google.com/p/metacello/issues/detail?id=108

I've tried to resolve this past two days (yep, not so easy for a first dive into Metacello :). If I understand:
'GoferBar-lr.branch.1'  should be  'GoferBar.branch-lr.1'

so I've changed this for all packages/branches everywhere (in MetacelloMonticelloResource, MetacelloConfigurationResource, GoferResource and tests).

Now a lot of tests are red because in MetacelloFetchingMCSpecLoader>>retryingResolvePackageSpecReferences: it cannot find GoferBar-lr.branch.2
Indeed in MetacelloTestConfigurationOfFoo generated by the tests there's
package: 'GoferBar' with: 'GoferBar-lr.branch.2';

This driving me mad: I cannot find how MetacelloTestConfigurationOfFoo is generated and when a spec with 'GoferBar-lr.branch.2' is created instead of 'GoferBar.branch-lr.2'.

Any hint ?




More information about the Pharo-project mailing list